A new species of Cerradomys ( Mammalia : Rodentia : Cricetidae ) from Central Brazil , with remarks on the taxonomy of the genus

Cerradomys is a Neotropical genus of cricetid rodents with seven recognized species, Cerradomys subflavus, C. maracajuensis, C. marinhus, C. scotti, C. langguthi, C. vivoi, and C. goytaca. Species of the genus are distributed throughout the open vegetation belt across South America, from northeastern and southeastern Atlantic coast of Brazil to eastern Paraguay and Western Bolivia. Here we describe a new species of Cerradomys from the state of Tocantins in Central Brazil, based on morphological, karyological and mitochondrial DNA analyses. This species is characterized by a medium body size and long tail, dense dorsal pelage, overall dorsal color gray olive lined with yellow, color of head and dorsum continuous, ventral body color slightly yellowish, skull with deep rostral depression, mesopterygoid fossa with long and wide sphenopalatine vacuities, presence of alisphenoid strut and of complex posterolateral palatal pits, and a unique chromosomal formula (2n = 60 and FNa = 74). Phylogenetic analyses based on cytochrome b sequences, including for the first time all known Cerradomys species, indicate that the new species is more closely related to C. scotti. The new species is found in sympatry with C. marinhus, while C. marinhus, C. scotti, and C. subflavus are found in sympatry (but not in syntopy) in one locality in the state of Minas Gerais. Finally, analysis of cytochrome b sequences indicates that C. subflavus and C. goytaca are very closely related genetically and might be conspecific. Alternatively, these results can also be explained by incomplete lineage sorting due to a recent speciation event.

The taxonomy of the members of this genus, however, is not yet fully explored and here we describe a new species from central Brazil, based on morphological, morphometric, and karyological evidence.Moreover, we also present a phylogenetic analysis for all Cerradomys species based on cytochrome b DNA sequence data, and discuss some taxonomic issues.

Origin of samples and karyotypic analysis
We collected Cerradomys specimens (Appendix 1) in nine Brazilian localities in three morphoclimatic domains (sensu AB 'SABER 2003) and in the ecotone between Cerrado and Amazonia: a) Cerrado domain: Tocantins state (1) Novo Jardim Chromosome preparations were obtained from shortterm cell cultures following de ANDRADE et al. (2004).Chromosomes were ordered according to morphology and decreasing size.For new karyotypes, several metaphases were captured and analyzed in the microscope, and five metaphases were mounted for each karyotyped specimen.Here only metacentric and submetacentric chromosomes were considered as biarmed for computation autosome fundamental number.

Phylogenetic analysis
DNA was isolated from tissue samples preserved in 100% ethanol following the protocol of SAMBROOK et al. (1989).Parcial cytochrome b gene (ca.801 bp) was amplified by PCR using primers L14724 (5'-CGAAGCTTGATATGAAAAACCATCGTTG-3' (IRWIN et al. 1991) andMVZ16 (SMITH &PATTON 1983).Amplicons were purified using the GFX PCR DNA and Gel Band Purification Kit (GE Healthcare, Brazil), and sequenced using the PCR primers.Sequencing was carried out with an ABI PrismTM 3730 automatic DNA sequencer.
Kimura two-parameter distances were estimated between haplotypes using PAUP* 4.0b10 (SWOFFORD 2001).For phylogenetic reconstructions, a DNA substitution model was selected using the software Modelgenerator, version 0.85 (KEANE et al. 2006) and the Bayesian information criterion (BIC).The GTR model of nucleotide substitution (RODRÍGUEZ et al. 1990), corrected for site-specific rate heterogeneity using gamma distribution with four classes (YANG 1994) was used in all model-based analyses.Cladistic parsimony (MP) analysis was performed using the heuristic search algorithm implemented by PAUP*, with 1000 replicates of random taxon addition and TBR branch swapping; nodal bootstrap values (FELSENSTEIN 1985) were calculated using 1000 pseudoreplicates.The Maximum-Likelihood (ML) trees were calculated using RaxML (STAMATAKIS 2006.).Bootstrap values for the likelihood analysis were calculated using 1000 pseudoreplicates.Bayesian analyses (BI) were performed using Markov chain Monte Carlo (MCMC) sampling as implemented in MrBayes 3. 1.2 (HUELSENBECK & RONQUIST 2001, RONQUIST & HUELSENBECK 2003).Uniform interval priors were assumed for all parameters except base composition, for which we assumed a Dirichlet prior.We performed four independent runs of 10,000,000 generations each, with two heated chains sampling for trees and parameters every 10,000 generations.The first 2,500,000 generations were discarded as burn-in, and the remaining trees were used to estimate posterior probabilities for each node.All analyses were checked for convergence by plotting the log-likelihood values against generation time for each run, using Tracer 1.6 (RAMBAUT & DRUMMOND 2007), and all estimates have effective sample sizes over 200.

Morphologic analysis
We studied skins, skulls and skeletons deposited in the mammal collection of Museu Nacional (MN), Universidade Federal do Rio de Janeiro, and in the mammal collection of the Laboratório de Biologia e Parasitologia de Mamíferos Silvestres Reservatórios (LBCE), IOC, Fiocruz, Rio de Janeiro, Brazil.A list of specimens examined and collecting localities is provided in Appendix 1.All measurements are expressed in millimeters (mm), except weight that is expressed in grams (g).The following external measurements were obtained from specimen tags or from wild caught specimens during fieldwork: total length (TL) or head-and-body length (HBL), length of the tail (LT), pinnae length (Ear), length of hind foot (HF), and weight (W).When necessary, head-and-body length (HBL) was obtained by subtracting length of tail from total length.
Eighteen cranial measurements (BONVICINO & WEKSLER 1998) were obtained with digital calipers to the nearest 0.01 mm: (GSL) greatest skull length; (CIL) condylo-incisive length, measured from the greater curvature of one upper incisor to the articular surface of the occipital condyle on the same side; (BOC) breadth of occipital condyles, measured from external border of the condyles; (BPB) breadth of palatal bridge, measured at the labial margin of maxillary bone across the third molars; (LD) length of diastema, from the crown of the first upper molar to the lesser curvature of the upper incisor on the same side; (LPB) length of palatal bridge, measured from the posterior border of the incisive foramen to the anterior border of the mesopterygoid fossa; (LIF) length of incisive foramina, greatest anterior-posterior dimension of one incisive foramina; (BIF) breadth of incisive foramina; (LR) length of rostrum, greatest dimension measured from the anterior border of the nasal ZOOLOGIA 31 (6): 525-540, December, 2014 bone to orbital fossa; (BR) breadth of rostrum, greatest dimension measured across the external border of the nasolacrimal capsules; (CH) cranial height, measured from dorsal surface of frontal to ventral surface of palatal bones, behind the third molar; (BB) breadth of braincase, measured across the smooth lateral surface of the braincase posterodorsal to the squamosal zygomatic processes; (LM) length of superior molars, crown length from M1 to M3; (BM1) breadth of M1, greatest crown breadth of the first maxillary molar across the paracone-protocone; (LIB) least interorbital breadth, least distance across the frontal bones; (LOF) length of orbital fossa, the greatest diameter of orbital fossa; (ZB) zygomatic breadth, greatest dimension across the squamosal root of zygomatic arches; (BZP) breadth of zygomatic plate.
Morphometric analyses of skull characters were performed for adult specimens (i.e., specimens with all teeth erupted and with at least minimal wear; OLIVEIRA et al. 1998); males and females were grouped due to lack of sexual dimorphism in adults (BRANDT & PESSOA 1994).Discriminant Analysis with estimation of canonical functions (STRAUSS 2010), using logarithmic-transformed data, were carried out for identifying patterns of morphometric variation within the genus, while univariate Analyses of Variance (SOKAL & ROHLF 1995) were employed for comparing the new species with the sister taxon C. scotti (see phylogenetic results).Greatest skull length (GSL) was excluded from discriminant analysis due to high frequency of missing values and high correlation with condylo-incisive length (r = 0.96).Sequential Bonferroni correction (RICE 1989) was used to adjust p-values for multiple contrasts in the ANOVA.Statistical analyses were performed with STATISTICA 7.0 (STATSOFT 2004).

Karyologic and molecular data
Karyotypic analysis of five C. akroai sp.nov.specimens showed 2n = 60 and FNa = 74 .The chromosome complement is composed by 11 pairs of biarmed chromosomes, one large sized, two median sized and 8 small sized, and 18 acrocentric pairs varying in size from large to small.The X chromosome is a large sized submetacentric and the Y chromosome a small sized acrocentric (Table I, Figs 2-11).Karyotypic analysis of five C. scotti specimens showed 2n = 58 and FNa = 70 (Table I, Figs 2-11).The nine specimens of C. marinhus showed 2n = 56 and FNa = 54, while karyotypic analysis of three C. goytaca specimens showed 2n = 54 and FNa = 62-63 (Table I, Figs 2-11).
Phylogenetic analyses of cytochrome b sequences, regardless of methodological approach (MP, ML, and BI), recovered the same general topology, with three main clades of Cerradomys (Fig. 12 The specimens identified as Cerradomys subflavus were never recovered in a monophyletic unit, as its terminals formed a polytomy with C. vivoi and C. goytaca (Fig. 12).In the parsimony analysis, C. vivoi formed a reciprocally monophyletic clade relative to the clade containing C. subflavus and C. goytaca, while in the ML and BI analyses the terminals of C. subflavus and C. goytaca are intermixed.
The new Cerradomys species exhibit the diagnostic characteristics of members of the genus (WEKSLER et al. 2006): interorbital region strongly convergent anteriorly, with well developed supraorbital crests (Figs 14-16); very long incisive foramina with lateral margins wider medially and antero-posterior margins sharp; stapedial foramen and posterior opening of alisphenoid canal vestigial or absent; squamosal-alisphenoid groove and sphenofrontal foramen absent; secondary anastomosis of internal carotid crossing the dorsal surface of pterygoid plate; capsular process of lower incisor developed.Dorsal pelage coarsely grizzled; tail longer than combined length of head and body, hind foot with small hypothenar pad, densely covered squamae distal to thenar pad, and conspicuous ungual tufts at bases of claws on dI-dV.
Cerradomys species exhibit morphological variation in integumental and cranial traits (see also PERCEQUILLO et al. 2008, TAVARES et al. 2011).The fur color varies among species, from buffy-yellow grizzled with dark-brown to orange-or red-buff grizzled with black; in C. subflavus, C. langguthi, C. vivoi, and   Given the observed differences in morphology and karyotype, and the results of the morphometric and phylogenetic analyses, we describe this form of Cerradomys as a new species.
Holotype.MN80491, an adult female specimen collected by Flávia Casado (original field number LBCE13509) in February 6, 2010.The holotype consists of skull, partial postcranial skeleton, and skin.A bone marrow suspension cells in Carnoy's fixative (methanol: acetic acid) and a liver tissue sample preserved in ethanol are housed at Laboratório de Biologia e Parasitologia de Mamíferos Silvestres Reservatórios, Instituto Osvaldo Cruz, FIOCRUZ, under the original field number LBCE13509.Cytochrome b DNA data were deposited in GenBank (accession number KP122219).
Paratypes.BRAZIL, Tocantins: Novo Jardim, males MN80488 (field number LBCE13447), MN80489 (LBCE 13468), MN80490 and long (21% of head and body length); pinnae rounded and small (Ear = 18 to 21 mm; 14% of head and body length), covered internally with short orange hairs and externally with yellow and dark hairs.Long and dense dorsum pelage gray olive lined with yellow, with many dark guard hairs, yellowish cover hairs, and gray and soft under hairs.Cover hairs long with a sub terminal yellow band; guard hairs sparse and long, with distal half entirely black or dark brown.General ventral color buffy or yellowish, slightly grizzled, and distinctively lighter than dorsal pelage, with hairs grayish-based and tipped with buffy or yellowish.Flanks bright yellow with dark guard hairs rare.Mystacial vibrissae long, reaching but not surpassing pinnae when laid back.Tail bicolored in the proximal half, covered with short and sparse brown hairs and scales on dorsal surface and unpigmented hairs and scales on ventral surface.Hind foot of medium size (24 to 32 mm), with dorsal surface white, covered with short, wholly white hairs.Ungual tufts sparse, shorter than or reaching the end of claws; ventral surface naked.
Diagnosis.Cerradomys akroai sp.nov.can be identified by the combination of the following characteristics: medium body size (HBL varying from 111 to 140 mm) and long tail (TL = 149 to 162 mm), dense dorsal pelage, overall dorsal color gray olive lined with yellow, head and dorsum with same color, ventral body color slightly yellowish, skull with deep rostral depression, mesopterygoid fossa with long and wide sphenopalatine vacuities with reduced basisphenoid penetration (4 presphenoid: 1 basisphenoid), alisphenoid strut present, basisphenoid foramen absent, deep palatal fossae (complex posterolateral palatal pits), and a unique chromosomal formula (2n = 60 and FNa = 74,. Description.Tail length longer than head and body (124% of head and body length); hind feet moderately narrow ZOOLOGIA 31 (6): 525-540, December, 2014 bital region narrow, converging anteriorly, with dorsolateral margins with sharp and well developed supraorbital crests.Braincase oblong, with prominent temporal crests.Zygomatic plate projected forward in lateral view.Incisive foramina long, with lateral margins concave and wider posteriorly; posterior margins sometimes reaching, but not extending between alveoli of upper first molars.Palate long and wide; posterolateral palatal pits numerous and complex, recessed in deep palatal fossae.Mesopterygoid fossa with anterior margin rounded or slightly acute, not reaching M3 alveoli; bony roof of mesopterygoid fossa perforated by large sphenopalatine vacuities, characterized as openings reaching the basisphenoid and wider than posterior expansion of presphenoid bone.Alisphenoid strut present.Postglenoid foramen large and nearly semicircular in shape separated from small or absent subsquamosal fenestra by a wide hamular process of squamosal.Mandible long; coronoid process large, falciform or triangular, nearly equal to condyloid process; superior notch shallow; angular process short, not surpassing the condyloid process posteriorly; inferior notch shallow; capsular process of lower incisor well developed.Cranial measurements of C. akroai and all other Cerradomys species are presented in Table II.
The upper incisors are opisthodont, with smoothly rounded enamel bands.The maxillary toothrows are parallel.Molars are bunodont and with labial flexi enclosed by a cingulum (Fig. 39).The first upper molar (M1) anterocone not divided into anterolabial and anterolingual conules, anteromedian flexus not present.The anteroloph is well developed, and can be joined or not with the anterocone by labial cingulum; the protostyle is absent; mesolophs are small on M1 and M2; the mesoloph is often joined to the paracone forming a single struc-ture.The paracone is connected by enamel bridge to the anterior moiety of the protocone.The protoflexus of M2 is poorly developed; the mesoflexus is present as single internal fossette; the paracone lacks an accessory loph.The third upper molar (M3) is reduced, and lacks a posteroloph, but has deep hypoflexus in unworn dentitions.The first lower molar (m1) anteroconid lacks an anteromedian flexid; an anterolabial cingulum is present on all lower molars; ectolophids are absent on m1 and m2; a small mesolophid is distinct on unworn m1 and m2, joined to the entoconid; a posteroflexid is well developed on m3.
Karyotype.Cerradomys akroai sp.nov.holotype shows a karyotype with 2n = 60 and FNa = 74, the highest diploid and fundamental number among Cerradomys species (Table I, Figs  2-11).See full description in results.Karyotypic data reinforces the uniqueness of C. akroai with respect to other congeneric species .This difference in the autossome complement would putatively lead to an infertile hybrid in the eventual mating of individuals between the two species (see discussion).
Comparisons.Despite the similarities shared by all congeneric species, C. akroai sp.nov.can be clearly distinguished from C. vivoi, C. langguthi, C. subflavus and C. goytaca by its external pelage coloration.In these four last species, the anterior half of dorsal head pelage is distinctively grayish to yellow-grayish and different from the remaining dorsal coloration, while in C. akroai sp.nov., C. maracajuensis, C. marinhus, and C. scotti the head coloration and dorsal body pelage coloration are similar.Cerradomys akroai sp.nov.differs from C. maracajuensis and C. marinhus by its longer, denser, and olive-brown dorsal pelage, vis-à-vis the yellow-brown dorsal pelage in the last two taxa, and by cranial features; the roof of mesopterygopid fossa is per-  Etymology.This species is named in honor to the Akroá, an extinct Amerindian people that occupied the Novo Jardim region until the XVIII and XIX centuries (Apolinário, 2006).

DISCUSSION
Cerradomys species are characterized by a remarkable karyotypic diversity (LANGGUTH & BONVICINO 2002, BONVICINO 2003, PERCEQUILLO et al. 2008), with several cases of intraspecific polymorphism (Table I).Cerradomys akroai sp.nov.possess the highest diploid and fundamental numbers within the genus, and is clearly diagnosed by diploid and/or fundamental autosome numbers, as well as morphology of autosomes and the sex chro-   akroai karyotype for comparison with C. scotti; nevertheless, the difference in the autossome complement between these two taxa is related to both diploid and fundamental autosome numbers; to derive one karyotype from another involves a series of chromosomal rearrangements including inversions and fusions or fissions.Even when these chromosomal rearrangements have little effect on hybrid fitness, they might reduce gene flow through the suppression of recombination, due to mechanical pairing problems; in this form, the rearrangements extend the effects of linked isolation genes, and thus facilitate speciation (REISEBERG 2001, KIRKPATRICK 2010).The effect of suppression of recombination offered by chromosomal rearrangements is more effective in species in which the number of chromosomal rearrangements is large, as is the case of C. scotti and C. akroai, where more than one chromosomal pair is involved in rearrangements.
In addition, no variation was found in diploid number among the 21 specimens of C. scotti karyotyped so far (Table I), suggesting that 2n is fixed; the FNa variation found in C. scotti (FNa = 70 or 72) is due to an pericentric inversion affecting a small chromosome pair, which is not related to the karyotype differences of the two species.Finally, C. scotti and C. akroai also differ in the morphology of the Y chromosome, a small sized acrocentric in C. akroai sp.nov., but a median sized biarmed chromosome in C. scotti.Karyological variation found in the fundamental autosome pairs of C. goytaca chromosomal complement in relation to the karyotype described by TAVARES et al. (2011) could be attributed to different interpretation related to the morphology of small acrocentric pairs, and also due to a pericentric inversion affecting small chromosomes (Fig. 11).Although changes in diploid number can be easily detected in rodent karyotypes, the fundamental number is not so easily evident due to different levels of condensation of small autosomes and also to different interpretations of biarmed chromosomes.Morphology of biarmed metacentric and submetacentric chromosomes, as well telocentric chromosomes, it not problematic.However, acrocentric chromosomes have their centromere close to the short arm telomere, and the short arm has a secondary constriction, of variable length with noncoding DNA (RIESEBERG 2001).In some preparations, small acrocentric chromosomes can appear as biarmed chromosomes, leading to different estimates of fundamental number.
Molecular data also support that the new karyomorph belongs to a distinct evolutionary lineage, sister to C. scotti.There are 37 fixed nucleotide differences between C. scotti and C. akroai, including one non-synonymous substitution.The average genetic distance (p) estimated for C. akroai and C. scotti is 5.7%, larger than the average distances between C. langguthi and C. subflavus (p = 4.7%), and much larger than between C. subflavus, C. goytaca, and C. vivoi (p = 1.7%; see below).Phylogenetic analyses also demonstrated the close relationship between C. maracajuensis and C. marinhus, and that a separate clade is formed by C. subflavus, C. vivoi, C. goytaca, and C. langguthi.All molecular analyses showed that C. subflavus and C. goytaca haplotypes are mixed, and that these forms are not reciprocally monophyletic, suggesting that and C. goytaca is a junior synonymous of C. subflavus.Although ML and BI analyses also showed that C. vivoi is not reciprocally monophyletic relative to C. subflavus and C. goytaca, MP analysis showed that the two groups are reciprocally monophyletic, and thus could be considered as distinct lineages.The reduced genetic distance estimate between C. goytaca and C. subflavus (average p = 0.6%) is smaller than the distances among any other species pair (average p = 2.3% between C. vivoi and C. subflavus; у 4.7% between all other species pairs); the estimate is in level with intraspecific variation of other Cerradomys species, such as C. maracujensis ( = 0.7%) and C. scotti ( = 0.6%).These analyses indicate that C. subflavus and C. goytaca might be conspecific; alternatively, the recovered phylogenetic pattern can be explained by a recent speciation event with incomplete lineage sorting.Further analyses with denser genetic and taxonomic sampling are necessary to establish a detailed relationship among these taxa.The morphometric differentiation of Cerradomys species is coincident with the phylogenetic arrangement; the three morphometric groups recovered by the first and second canonical functions corresponded to the three clades: (C. vivoi, C. subflavus, C. goytaca, C. langguthi), (C. scotti, C. akroai), and (C. marinhus, C. maracajuensis).Cerradomys akroai is more similar to its sister species C. scotti than to any other congeneric species.They differ, however, in at least five variables, mainly in height of skull (CH), but also in orbital region (LZIG, CORB), rostrum (LROS), and palate (PPAL).The relationship between morphometric and genetic differentiation of this genus also deserves a more detailed analysis.
Cerradomys akroai, together with C. scotti, are endemic of open vegetation domain, occupying the central area of the genus geographic range (Fig. 1).Interestingly, both species are sympatric with C. marinhus in their respectively type localities.Cerradomys scotti is also sympatric with C. maracajuensis in the Brazilian states of Mato Grosso and Mato Grosso do Sul, Sapucay in Paraguay, and Santa Cruz in Bolivia (PERCEQUILLO et al. 2008).We also report for the first time the sympatry (but not syntopy) of three species -C.marinhus, C. scotti, and C. subflavus -in Uberlândia, Minas Gerais (Appendix 1).
Cerradomys akroai and C. scotti are more associated with open vegetation formation of the Cerrado domain, while C. maracajuensis, C. subflavus, C. vivoi, C. goytaca, and C. langguthi are mainly associated to forest formations that occur within the Cerrado, such as gallery forest, and in ecotones or in the Atlantic forest, while C. marinhus is more often found in the grass marshes with buriti palms called "veredas" (EITEN 1983) and flooded forests.
Our study confirms that Cerradomys is a speciose taxon with a core distribution in the open vegetation belt of Eastern South America and that could be a model for the study of the biogeography of the region.Some Cerradomys species distributional limits are shaped by the Rio São Francisco, such as C. langguthi and C. subflavus/C.vivoi, a pattern similar to that found for other open vegetation belt taxa, such as Thrichomys (NASCIMENTO et al. 2013), Gracilinanus agilis (Burmeister, 1854) (FARIA et al. 2013), and Calomys expulsus (Lund, 1841) (NASCIMENTO et al. 2011).Other processes could have played a role on evolutionary history of the Cerradomys, such as Pleistocene climatic oscillations that lead to reduction of open vegetation areas during interglacial periods.Our understanding of the biogeographic processes that shaped the distributional limits for Cerradomys species still awaits a more extensive study with denser geographic sampling.
): I) C. akroai sp.nov.and C. scotti; II) C. maracajuensis and C. marinhus; and III) C. langguthi, plus a clade with the three remaining species of the genus (C.subflavus, C. vivoi, C. goytaca).The clade (C.maracajuensis, C. marinhus) is found as sister group to the other two Cerradomys clades in the ML and BI, but the consensus of 8 most parsimony trees reveals a polytomy involving the three main clades (results not shown).Cerradomys akroai sp.nov.and C. scotti are found with weak or moderate support in the different phylogenetic approaches (MP bootstrap 79%, ML bootstrap 61; BI posterior probability 0.80).

Figure 12 .
Figure 12.Phylogenetic relationships of Cerradomys species based on maximum likelihood (ML) analysis of Cytochrome b sequences using the GTR+G(4) model.Bayesian (BI) and parsimony (MP) analyses recovered similar trees.Nodal support indices are shown as shaded pie diagrams at nodes.Colors indicate level of support: for MP and ML, black indicates bootstrap values above 85%, gray indicates values between 50 and 85%, and white indicates values below 50%.For BI posterior probability, black indicates values above 0.95, while white indicates values below 0.95.The tree is rooted with Necromys and Rhipidomys.Average genetic distances (p) between sister clades are shown below nodes; intraspecific nucleotide diversity (, in italic) are shown above species nodes.
and wide sphenopalatine vacuities, exposing the presphenoid and basisphenoid in C. akroai sp.nov.and C. scotti, while C. maracajuensis and C. marinhus are characterized by shorter vacuities restricted to the presphenoid or by a completely ossified roof of the mesopterygoid fossa.Finally, C. akroai can be distinguished from all species except C. scotti, by the presence of the alisphenoid strut and the strongly dorsoventrally bicolored tail in proximal half.Cerradomys akroai sp.nov. is morphologically more similar to C. scotti than to any other Cerradomys species, but can be distinguished from C. scotti by its dorsal body color, which is darker in C. akroai sp.nov., and by the following cranial differences (Figs17-22): 1) sphenopalatine vacuities in C. akroai sp.nov.with a less posterior penetration into the basisphenoid (extension of vacuities: 4 presphenoid: 1 basisphenoid) than in C. scotti (2 presphenoid: 1 basisphenoid); 2) conspicu-ous foramen that perforates the basisphenoid in C. scotti, but is apparently absent in C. akroai sp.nov.; 3) and posterolateral palatal pits deeper in C. scotti than in C.akroai sp.nov.
The scores' range of the last species overlapped with the other groups, and C. maracajuensis was the least correctly classified species in discriminant functions, with only 63% of correct cases; C. vivoi also had a low classification score, with 62% of correct cases.Three species (C.scotti, C. goytaca, and C. akroai sp.nov.) had all specimens correctly classified, while predicted classification of remaining species varied from 78% (C.subflavus; 1 specimen misclassified as C. langguthi and 1 as C. vivoi) to 93% (C.marinhus; 2 specimens misclassified as C. maracajuensis).

Table I .
Karyotypic data of Cerradomys.Brazilian states acronyms as in Fig.1.

Table II .
Descriptive statistics of cranial measurements of Cerradomys species: sample size (when different from first row)_average ± standard desviation (minimum-maximum).